Analysis Methods | Variant Calling | Somatic Variant Caller

Variant Calling

Developed by Illumina, the somatic variant caller identifies variants present at low frequency in the DNA sample.

The somatic variant caller identifies SNPs in 3 steps:

▶ Considers each position in the reference genome separately
▶ Counts bases at the given position for aligned reads that overlap the position
▶ Computes a variant score that measures the quality of the call using Poisson model.

Variants are first called for each pool separately. Then, variants from each pool are compared and combined into a single output file. If a variant meets the following criteria, the variant is marked as PASS in the variant call (VCF) file:

▶ The variant is present in both pools
▶ Has a cumulative depth of 1000 or an average depth of 500x per pool
▶ Has a variant frequency of ≥ 3% as reported in the merged VCF file

A locus for a mutation or reference is classified as a no call under the following conditions:

▶ The variant frequency is near the signal noise level between 1% and 2.6%
▶ The variant quality is < Q30
▶ The depth is < 500
▶ Significant strand bias is detected
▶ The indel occurs in a homopolymer region